<MANDATORY_RULES>
These rules apply ALWAYS, without exception, for EVERY action:
@CLAUDE/CLAUDE.subagents.md
BEFORE EVERY ACTION check:
- Does this require user interaction? -> Main Agent OK
- Is there a specialized agent for this? -> USE IT!
- No fitting agent? -> general-purpose subagent
- 2+ independent tasks? -> Hydra MANDATORY
NEVER work directly when a subagent exists.
@CLAUDE/CLAUDE.git.md
ALWAYS:
- No AI attribution (Co-Authored-By, "Generated with Claude")
- Never delete local files without explicit user confirmation
@CLAUDE/CLAUDE.ai-traces.md
ALWAYS:
- No curly quotes, em-dashes, smart ellipsis
- No AI phrases in code, comments, or logs
- No emojis in code
</MANDATORY_RULES>
@CLAUDE/CLAUDE.honesty.md @CLAUDE/CLAUDE.language.md @CLAUDE/CLAUDE.planning.md @CLAUDE/CLAUDE.prompt-intervention.md @CLAUDE/CLAUDE.prompt-injection.md @CLAUDE/CLAUDE.philosophy.md
<situational_rules> Read these files only when relevant:
Project Setup:
CLAUDE/CLAUDE.naming.md- when naming projects, repos, or packagesCLAUDE/CLAUDE.structure.md- when modifying repo structure or adding filesCLAUDE/CLAUDE.environment.md- when setting up .env or config
Security & Legal:
CLAUDE/CLAUDE.security.md- when adding dependencies or handling secretsCLAUDE/CLAUDE.legal.md- when adding LICENSE, privacy policy, or impressum
Code Quality:
CLAUDE/CLAUDE.linting.md- when setting up or configuring ESLintCLAUDE/CLAUDE.formatting.md- when setting up or using PrettierCLAUDE/CLAUDE.build.md- when setting up Vite, Vitest, or build toolsCLAUDE/CLAUDE.testing.md- when writing or running testsCLAUDE/CLAUDE.error-handling.md- when implementing error handling or logging
Workflow:
CLAUDE/CLAUDE.versioning.md- when bumping versions or releasingCLAUDE/CLAUDE.branching.md- when creating branches or PRsCLAUDE/CLAUDE.code-review.md- when reviewing codeCLAUDE/CLAUDE.documentation.md- when writing docs or comments
Optional (end-user apps only):
CLAUDE/CLAUDE.accessibility.md- when building user-facing web UI </situational_rules>